csuter
commented
4 years ago - flatten out the distribution of delta_t; no longer a JointDistributionNamed, just a FiniteDiscrete
- replaced OneHotCategorical with a Categorical (sample needs an expand dims after)
- fix bug -- were using zeros instead of int max's for upper bounds
- thread dtypes through various functions
- inlined some arg-free distributions in JointDistributionNamed
- minor reformatting to appease Chris Suter's formatting compulsions
- pprint in tests; fix test data paths (weren't working for Chris Suter's setup)